What to Expect From a Data Cleaning Project

You've decided your CRM needs professional cleaning. Maybe the bounce rates forced the issue. Maybe a CRM migration is coming. Maybe your sales leader finally asked why the pipeline report doesn't add up.

Whatever the trigger, you're about to hand your data to someone else, and most people doing this for the first time have no idea what a normal project looks like. So here's the whole thing, stage by stage: what happens, how long each part takes, what it should spend, and the warning signs that you picked the wrong provider.

You send files out, verified and corrected files come back, and your team keeps selling the entire time.

Stage 1: Scoping (Day 1)

Before any work starts, the provider needs to understand your database. A good scoping conversation covers:

• Database size: How many contact, account, and lead records?
• Known issues: Duplicates? Bouncing emails? Inconsistent job titles? Missing fields?
• Services needed: Deduplication only? Full cleaning? Cleaning plus enrichment?
• CRM platform: Salesforce, HubSpot, Dynamics, or other?
• Custom requirements: Specific fields to standardize? Particular merge rules? Records to exclude?

After discovery, the provider should document eligible records, included checks, exception rules, review responsibilities, delivery format, schedule, usage terms, and the project quote. Ask how change requests will be handled. Ambiguity at this stage becomes scope creep later.

One more thing worth settling now: ask how they handle records they can't fix. The right answer involves flagging and reporting. The wrong answer involves quiet deletion.

Stage 2: Data Export (Days 1-2)

You'll export your data from your CRM and send it to the provider. Most providers accept CSV or Excel files. A few can connect directly via API.

What to export:

• Contacts/Leads: fields including name, email, phone, title, company, address
• Accounts/Companies: Company name, industry, size, address, website
• Record IDs: So cleaned records can be matched back to your CRM on import instead of creating a second copy of everything

This stage is also where security gets real. You're sending customer data outside your walls, so confirm the transfer method (SFTP or an encrypted share, rarely plain email attachments for large files), ask how long the provider retains your file after delivery, and get a data processing agreement if your records include EU contacts.

Stage 3: Processing (Days 2-6)

This is where the cleaning happens, and it's the part you see least. A full-scope project runs through these steps in order:

Deduplication

Fuzzy matching proposes records that may represent the same person or company despite spelling, formatting, or completeness differences. Define field-survival rules, preserve source identifiers, and route ambiguous clusters to human review before any merge is approved.

Email Validation

Each email address is checked at the SMTP level and categorized: valid, invalid (would hard bounce), risky (domain with domain-wide acceptance that accepts everything), or role-based (info@, admin@). Invalid addresses get flagged for removal. Risky ones get flagged for your judgment, since some uncertain addresses are perfectly reachable and some are black holes.

Phone Verification

Phone numbers run against carrier databases. Disconnected numbers, landlines, and fax lines get flagged separately from active direct dials and mobiles, so your reps stop burning call blocks on numbers that ring nowhere.

Field Standardization

Job titles, company names, addresses, and industry values get normalized to consistent formats. "VP of Sales" and "Vice President, Sales" become one value. "CA", "Calif.", and "California" become one value. This is the step that quietly fixes your broken lead routing and segmentation, because those systems fail on inconsistency more than on absence.

Enrichment (if included)

Missing fields get filled from external sources: emails, phones, titles, company size, industry. This is a different operation from cleaning, with its own expectations around match rates. Our data enrichment guide covers how to judge that part of the quote.

Stage 4: Review (Days 6-7)

Before you import anything, the provider should deliver four things:

1. The cleaned data file formatted for your CRM's import tool
2. A change log showing what was modified, added, or flagged on each record
3. Summary statistics: duplicates found and merged, emails validated versus flagged, phones verified versus disconnected, fields standardized, and fill-rate improvements if enrichment was included
4. Recommendations for preventing the same issues from recurring

Then do your own check. Review a representative sample of approved, rejected, and ambiguous records; confirm field mappings and related records before importing the full delivery.

Stage 5: Import (Days 7-8)

Import the cleaned data back into your CRM. Most providers deliver files formatted for your specific import tool (Salesforce Data Loader, HubSpot native import, and so on).

• Back up your CRM first. Generally have a rollback point.
• Import a small, reversible batch first and verify mappings, relationships, automation, and rejected rows before continuing.
• Match on record IDs. Update existing records rather than creating new ones.
• Test your automation after import. Lead scoring, routing, and sequences need to work with the newly standardized values.

Records your team created while the project ran are unaffected; they just weren't part of this pass. They get covered in the next one.

What Results Should You See Afterward?

After importing, compare bounce status, segment counts, routing exceptions, duplicate candidates, and pipeline reports with the dated baseline. Attribute improvement only where the same definition and reporting window were used.

A cleaning project is a dated pass, not a permanent state. Measure changes against the baseline, identify which sources and workflows created new exceptions, and set the next review from observed change. If you want to grade the database before hiring anyone, start with the CRM data quality checklist.

Red Flags When Hiring a Provider

Warning signs include missing status definitions, no exception file, unreproducible totals, and a provider unwilling to show how accepted and rejected records were sampled.

The change log question is the fastest filter. Ask "will I get a record-level log of each change?" If the answer is anything other than yes, you'll have no way to audit what happened to your data, and no way to undo it.

What should the handoff include?

The delivery should separate changed, unchanged, unresolved, and excluded records. Include the ruleset version, source fields, proposed values, review status, and a summary of exceptions. A rollback copy and import order help the CRM owner test safely. Acceptance is clearer when both teams can trace a changed value to the rule or evidence that produced it.

Before sign-off, reconcile row counts from intake through delivery and inspect the unresolved queue. Record who approved the import, which records remain excluded, and how a future reviewer can reproduce the checks. Store the dated input, output, and exception files together under the same project identifier so later audits compare the correct versions.

Frequently Asked Questions

How long does a data cleaning project take?

The written scope sets the schedule after reviewing record count, systems, requested checks, match rules, protected fields, exception volume, approval steps, and delivery format.

How much does a data cleaning project cost?

Managed cleaning is scoped around record volume, requested services, exception handling, manual review, and delivery requirements. Expect the written proposal to define the unit of work, included steps, change process, schedule, and project quote before authorization.

What should a data cleaning deliverable include?

Four things: the cleaned file formatted for your CRM's import tool, a change log showing what was modified on each record, summary statistics (duplicates merged, emails validated, phones verified, fields standardized), and recommendations for keeping the data clean. A provider who hands you a file with no change log is asking you to trust them blind.

How do I prepare my CRM for a data cleaning project?

Export contacts, accounts, and leads as CSV files with fields, including record IDs so cleaned rows can be matched back on import. Note custom fields and picklist values, and flag which records are most critical (open deals, recent leads) so the provider can prioritize them.

Will a data cleaning project disrupt my sales team?

The processing itself happens on an exported copy, so reps keep working in the CRM the whole time. The only coordination point is the import window. Schedule it for a low-activity period, freeze bulk edits for a day, and records created during processing directly get picked up in the next cleaning pass.

What happens to records that can't be fixed?

They come back flagged, rarely silently deleted. A contact whose email fails validation and whose company no longer exists gets marked for your review, and you decide whether to archive or keep it. Good providers report the unfixable segment because it tells you where your data sources are failing.